econstor Make Your Publications Visible.

Transcription

1 econstor Make Your Publcatons Vsble. A Servce of Wrtschaft Centre zbwlebnz-informatonszentrum Economcs Neundorf, Anja; Adams, James F. Workng Paper The mcro-foundaton of party competton and ssue ownershp: The recprocal effects of ctzens' ssue salence and party attachments SOEPpapers on Multdscplnary Panel Data Research, No. 692 Provded n Cooperaton wth: German Insttute for Economc Research (DIW Berln) Suggested Ctaton: Neundorf, Anja; Adams, James F. (2014) : The mcro-foundaton of party competton and ssue ownershp: The recprocal effects of ctzens' ssue salence and party attachments, SOEPpapers on Multdscplnary Panel Data Research, No. 692, Deutsches Insttut für Wrtschaftsforschung (DIW), Berln Ths Verson s avalable at: Standard-Nutzungsbedngungen: De Dokumente auf EconStor dürfen zu egenen wssenschaftlchen Zwecken und zum Prvatgebrauch gespechert und kopert werden. Se dürfen de Dokumente ncht für öffentlche oder kommerzelle Zwecke vervelfältgen, öffentlch ausstellen, öffentlch zugänglch machen, vertreben oder anderwetg nutzen. Sofern de Verfasser de Dokumente unter Open-Content-Lzenzen (nsbesondere CC-Lzenzen) zur Verfügung gestellt haben sollten, gelten abwechend von desen Nutzungsbedngungen de n der dort genannten Lzenz gewährten Nutzungsrechte. Terms of use: Documents n EconStor may be saved and coped for your personal and scholarly purposes. You are not to copy documents for publc or commercal purposes, to exhbt the documents publcly, to make them publcly avalable on the nternet, or to dstrbute or otherwse use the documents n publc. If the documents have been made avalable under an Open Content Lcence (especally Creatve Commons Lcences), you may exercse further usage rghts as specfed n the ndcated lcence.

2 SOEPpapers on Multdscplnary Panel Data Research SOEP The German Soco-Economc Panel Study at DIW Berln The Mcro-Foundaton of Party Competton and Issue Ownershp: The Recprocal Effects of Ctzens Issue Salence and Party Attachments Anja Neundorf and James Adams

3 SOEPpapers on Multdscplnary Panel Data Research at DIW Berln Ths seres presents research fndngs based ether drectly on data from the German Soco- Economc Panel Study (SOEP) or usng SOEP data as part of an nternatonally comparable data set (e.g. CNEF, ECHP, LIS, LWS, CHER/PACO). SOEP s a truly multdscplnary household panel study coverng a wde range of socal and behavoral scences: economcs, socology, psychology, survey methodology, econometrcs and appled statstcs, educatonal scence, poltcal scence, publc health, behavoral genetcs, demography, geography, and sport scence. The decson to publsh a submsson n SOEPpapers s made by a board of edtors chosen by the DIW Berln to represent the wde range of dscplnes covered by SOEP. There s no external referee process and papers are ether accepted or rejected wthout revson. Papers appear n ths seres as works n progress and may also appear elsewhere. They often represent prelmnary studes and are crculated to encourage dscusson. Ctaton of such a paper should account for ts provsonal character. A revsed verson may be requested from the author drectly. Any opnons expressed n ths seres are those of the author(s) and not those of DIW Berln. Research dssemnated by DIW Berln may nclude vews on publc polcy ssues, but the nsttute tself takes no nsttutonal polcy postons. The SOEPpapers are avalable at Edtors: Jürgen Schupp (Socology) Gert G. Wagner (Socal Scences, Vce Dean DIW Graduate Center) Conchta D Ambroso (Publc Economcs) Dens Gerstorf (Psychology, DIW Research Drector) Elke Holst (Gender Studes, DIW Research Drector) Frauke Kreuter (Survey Methodology, DIW Research Professor) Martn Kroh (Poltcal Scence and Survey Methodology) Freder R. Lang (Psychology, DIW Research Professor) Hennng Lohmann (Socology, DIW Research Professor) Jörg-Peter Schräpler (Survey Methodology, DIW Research Professor) Thomas Sedler (Emprcal Economcs) C. Katharna Speß (Emprcal Economcs and Educatonal Scence) ISSN: (onlne) German Soco-Economc Panel Study (SOEP) DIW Berln Mohrenstrasse Berln, Germany Contact: Uta Rahmann soeppapers@dw.de

4 The Mcro-Foundaton of Party Competton and Issue Ownershp: The Recprocal Effects of Ctzens Issue Salence and Party Attachments Anja Neundorf 1 (Unversty of Nottngham) James Adams (Unversty of Calforna, Davs) [Presented at the Annual Meetng of the Amercan Poltcal Scence Assocaton, Panel Partcpaton, Representaton and Dfferent Forms of Ext ; Washngton, DC, August 31 st 2014.] Abstract Whle prevous research on the recprocal effects of ctzens ssue atttudes and ther party support emphasze ctzens ssue postons, poltcal competton revolves equally around ssue salence,.e., debates over whch ssue areas poltcal partes should prortze. Usng mult-wave panel data from Germany and Great Brtan, we analyze the recprocal effects of ctzens ssue salence and ther party support over the perod , and we conclude that ctzens ssue prortes both nfluence and are nfluenced by ther party attachments, and, moreover, that these effects are lnked to partes long-term polcy emphases as artculated n ther electon manfestos. Ths effect s strongest among supporters of a small ssue-orentated nche party, the Greens. 1 Emal-addresses: anja.neundorf@nottngham.ac.uk and jfadams@ucdavs.edu.

5 Introducton The study of how ctzens ssue concerns nfluence ther votes has prompted two, related, research agendas. The frst, whch s emphaszed by spatal modelers and by many behavoral researchers, s that partes offer competng ssue postons to voters whose decsons turn on the match between ther own polcy belefs and partes postons (see, e.g., Downs 1957; Pardos- Prado and Dnas 2010). Ths postonal perspectve prompts scholars to emphasze the electoral benefts poltcal partes gan by presentng polces that reflect publc opnon. The second perspectve s that partes compete by emphaszng dfferent ssues pertanng to domans on whch they enjoy ssue ownershp, n the sense that voters beleve the focal party s well-equpped to address the ssue (e.g., Budge and Farle 1983; Petrock 1996; Green and Hobolt 2008). Ths perspectve mples that partes talk past each other, wth partes that enjoy reputatons for wse economc stewardshp emphaszng the economy, partes wth strong reputatons for fghtng crme emphaszng the crme ssue, and so on. Wth respect to postonal ssue votng, a lvely emprcal lterature nvestgates whether ctzens choose partes on the bass of ther polcy postons or whether partes recprocally cue ther pre-exstng partsans to adopt the party s postons (see, e.g., Evans and Andersen 2004; Carsey and Layman 2006; Dancey and Goren 2010). To date, however, we are unaware of parallel research that evaluates the recprocal nfluences of ctzens party support and ther ssue salence. The lnk between these two serves as the mcro foundaton of ssue ownershp and party competton. In recent years, the focus has shfted to the condtonal effect of ssue salence, gven research that concludes that ssue ownershp only affects the decson of those voters who thnk that the focal ssue s mportant (Belanger and Megud 2008: 479; see also Anand and Krosnck, 2003; Pardos-Prado et al. 2013). However, we stll do not know whether ctzens ssue salence represents an endogenous factor that drves ther partsanshp, or vce versa. That s the ssue we address here. Specfcally, we analyze a 26-wave German panel and n supple-

6 2 mentary analyses we extend our focus to Brtan to evaluate the extent to whch ctzens ssue salence wth respect to the natonal economy, the envronment, crme, and mmgraton nfluenced ther subsequent party support a partsan updatng effect and the extent to whch, recprocally, ctzens party support shaped ther subsequent ssue prortes, an ssue cueng effect. We advance three arguments about the recprocal relatonshps between ctzens ssue concerns and ther party support. Frst, buldng on prevous fndngs that ctzens ssue salences are more malleable than ther ssue postons (e.g., Page and Shapro 1992), we argue that ctzens ssue concerns shape ther party support and ther party support shapes ther ssue prortes (the recprocal effects hypothess). Ths mples that ssue salence s endogenous to partsanshp (Walgrave et al. 2014). Second, we argue that the drecton of the effects we dentfy that, for nstance, envronmental concerns drve ctzens towards green partes but away from the centerrght partes that prortze economc growth over envronmental protecton are ted to the partes long-term ssue emphases, as artculated n ther electon manfestos (the party manfestos hypothess). Thrd, we argue for a Green Party exceptonalsm hypothess, that mass-elte lnkages wth respect to ctzens ssue salence are far stronger wth respect to the Green Party an ssue-orented nche party than wth respect to the manstream German partes, the Socal Democrats (SPD), the Chrstan Democrats (CDU/CSU), and the Free Democrats (FDP). Ths argument hghlghts the dfferences between catch-all partes that emphasze a wde range of ssues and nche partes that emphasze specfc ssues on whch they enjoy ownershp. We fnd support for all three hypotheses, and beleve our results are mportant for several reasons. Frst, our fndngs n support of the recprocal effects hypothess pertan to the argument that ssue ownershp affects the votng decsons of only those ndvduals who prortze the ssue (Belanger and Megud 2008: 477). However, we demonstrate here that ctzens ssue prortes and ther party support recprocally nfluence each other. Second, our emprcal support for the party manfestos hypothess, that ndvdual-level partsan updatng and ssue cueng effects reflect the ssue prortes German partes artculate n

7 3 ther manfestos, llustrates how partes manfestos medate ther ownershp of dfferent ssue domans, and also partes abltes to shape ther supporters ssue salence (e.g., Walgrave et al. 2014). Our study of ctzens reactons to partes manfesto-based ssue emphases thereby extends earler studes on the electoral effects of partes manfesto-based postons (see, e.g., Adams and Somer-Topcu 2009). Below we dscuss the mplcatons of our fndngs for German partes electon strateges. Thrd, our fndngs n support of the Green Party exceptonalsm hypothess extend the research of Megud (2008) and Spoon (2011), who conclude that mass-elte lnkages nvolvng nche partes specfcally green and radcal rght partes dffer from those nvolvng manstream partes. Megud and Spoon hghlght dfferences n the types of ssues nche partes emphasze, and we extend ths perspectve to show that the German Greens not only emphasze dfferent ssues than manstream partes, but that the Greens are also vastly more successful n attractng support on the bass of the ssues they emphasze, and n drectng ther supporters attenton towards these ssues (and away from the ssues they de-emphasze). The Recprocal Relatonshps between Ctzens Issue Salence and ther Party Support: Hypotheses In the Unted States, the debate over the recprocal nfluences of ctzens partsanshp and ther ssue postons has ntensfed n recent years. The conventonal wsdom of the 1970s and 1980s that mass partsanshp was weakenng and was largely drven by other poltcal evaluatons, ncludng polcy-based consderatons (Jackson 1975; Page and Jones 1979; Forna 1981) has been challenged by research that documents strengthenng partsan tes that ncreasngly nfluence vote choce (e.g., Green et al. 2002; Hetherngton 2001). Scholars have extended ths debate by analyzng the recprocal nfluences of ctzens ssue postons and ther partsanshp across dfferent domans (e.g. Goren 2005; Carsey and Layman 2006; Hghton and Kam 2011; Dancey and Goren 2010). These studes conclude that partsanshp nfluences Amercan ctzens ssue pos-

8 4 tons and poltcal values. By contrast, studes conclude that European ctzens partsanshp s less central to ther self-mages than are ther ssue postons, n partcular that partsanshp s more volatle n Europe than n the U.S., whch mples that partsanshp may not represent a salent dentty to Europeans (e.g., Clarke et al. 2009). Emprcal research by Mlazzo et al. (2012) supports ths perspectve that European voters are Downsan n that ther ssue postons nfluence but are not nfluenced by ther party attachments. Although the studes summarzed above support the prmacy of European ctzens ssue postons vs-à-vs ther partsanshp, there are reasons to doubt such a undrectonal causal relatonshp nvolvng ctzens ssue prortes. In partcular, Budge and Farle (1983) advance a salency theory of party competton, whch they support va emprcal analyses of partes electon manfestos, that poltcal partes n Europe (and elsewhere) selectvely emphasze ssues on whch they have a competence advantage rather than drectly engagng wth rval partes ssues (see also Klngemann et al. 1994; Budge et al. 2001; van der Brug 2004). In ths regard, scholars have developed the theory of ssue ownershp, that voters wll support the party that owns the ssue the voter prortzes, so that a voter preoccuped wth crme wll support (all else equal) the party wth a postve competence mage wth respect to law and order, voters preoccuped wth the economy wll support partes wth postve reputatons for economc stewardshp, and so on (Petrock 1996; Petrock et al. 2003). Emprcally, research by Hobolt et al. (2009) analyzes the lnk between party leaders ssue emphases and rank-and-fle ctzens ssue concerns n Brtan and Denmark, and documents that governng party leaders annual speeches shape the dversty of these publcs ssue prortes. Moreover, Green and Hobolt (2008) show that as partes converge deologcally, competence consderatons that are central to ssue ownershp become more salent to voters. Specfcally, extensve research on ssue-based party competton argues that ctzens ssue prortes shape ther party support (e.g., Petrock 1996; Klngemann et al. 1994; Clarke et al. 2009; Pardos-Prado et al. 2013).

9 5 The relatonshp between ssue concerns and party support s medated by the salency of these ssues to the voter (Budge and Farle 1983), whch may reflect her values or the socal group wth whch she dentfes (Krosnck 1990; Belanger and Megud 2008). Hence we expect that the more mportant an ssue s to the voter, the more lkely that she wll support the party that has the hghest competence mage wth regard to ths ssue. We label ths a partsan updatng effect. However, a recent study by Walgrave and hs colleagues (2014) presents evdence that party dentfcaton constrans ctzens abltes to receve and accept party messages. Ths s n lne wth the argument that partsanshp functons as a perceptual screen (Campbell et al. 1960; Converse 1964). Based on ths, we can alternatvely expect that ctzens party support shapes ther subsequent ssue prortes, whch we label an ssue cung effect. These consderatons motvate our frst hypothess: H1 (The Recprocal Effects Hypothess): Ctzens ssue salences both nfluence and are nfluenced by ther partsan afflatons. Next we dstngush two dfferent party-level factors that we beleve condton the recprocal effect hypothess frstly, party type as of nche versus manstream partes and secondly, partes ssue ownershp as transmtted through party manfestos. Recprocal effects of ssues and partsanshp: The case for Green Party exceptonalsm Our second hypothess s motvated n part by the emprcal work of Ktschelt (1994), Tarrow (1989), and Adams et al. (2006, 2012), who analyze the atttudes of poltcal eltes belongng to nche partes, specfcally small, ssue-focused partes such as green and radcal rght partes, along wth the characterstcs of these partes rank-and-fle supporters (Wagner 2012). These studes report fndngs that nche party eltes, actvsts, and rank-and-fle supporters place greater emphass on ssue debates than do ther counterparts from manstream partes, who frequently emphasze ther party eltes leadershp abltes and superor competence to govern. In partcular, Ktschelt (1994) fnds that nche partes core supporters are strongly polcy-orented and ac-

10 6 tvely oppose deologcal compromses by ther party s eltes. Second, and related, nche party supporters are more lkely to perceve and react to ther preferred party s polcy shfts than are manstream partsans. In ths regard Adams et al. (2006) demonstrate that unlke manstream partes, nche partes vote shares declne precptously when these partes moderate ther polcy postons, and, moreover, that nche party supporters are more poltcally-engaged than are manstream partes supporters fndngs whch suggest that nche party supporters dsproportonately montor ther preferred party s postons. Although these studes pertan to postonal ssues, to the extent that these patterns extend to ssue salence they mply that mass-elte lnkages should be dsproportonately strong wth respect to nche partes. Hence we expect ctzens ssue prortes and ther party support to exert stronger recprocal effects wth respect to nche partes. Here we follow Wagner s (2012: 845) defnton of nche partes as partes that compete prmarly on a small number of non-economc ssues. Gven that the German Greens a small party that prortzes envronmental ssues, whch we dscuss n more detal below s the only promnent nche party n Germany (Poguntke 1993), the studes summarzed above suggest the hypothess: H2 (The Green Party Exceptonalsm Hypothess): The recprocal nfluences of German ctzens ssue salence and partsanshp wll be stronger wth respect to the Greens than wth respect to the manstream partes. Issue effects and the role of partes electon manfestos The dscusson of voter reactons to partes ssue emphases rases the queston: How do partes communcate ther ssue emphases to the publc? Here we focus on the ssues that partes emphasze n ther natonal electon manfestos. The lengthy ntra-party dscussons and consultatons nvolved n composng these documents, along wth the extensve meda coverage of manfestos, testfy to ther central role n natonal electon campagns (see Budge et al. 2001). In addton, Adams et al. (2011) report ntervews wth party eltes from Germany, the Netherlands, and Austra, n whch these poltcans consstently assert that ther party makes determned efforts to

11 7 campagn based on ts electon manfestos, whle Baumgartner et al. (2009) report that the ssues partes prortze n ther electon manfestos are postvely correlated wth the ssue domans they prortze n other venues ncludng parlamentary debates, legslatve behavor, and government budgets. Assumng that electon manfestos are an mportant venue n whch partes artculate ther ssue emphases, and also that these manfesto-based emphases reflect the ssues partes emphasze n other venues, we next need to connect party manfestos to ctzens polcy salence and partsanshp. We argue that the recprocal relatonshp between ssue prorty and partsanshp (the recprocal effects hypothess) should be moderated by partes polcy emphases, namely that partes ssue emphases are the causal mechansm that drve the ndvdual-level partsan updatng and ssue cueng effects that we estmate. Based on Zaller s (1992) receve-accept-sample model, we expect partsans that receve nformaton about ther preferred party s ssue prortes to be more lkely to accept ths nformaton, as they have general trust n ths party (see also Walgrave et al. 2014). Recprocally, we expect ctzens who are concerned wth a specfc poltcal ssue to support partes whch emphasze ths ssue n ther manfesto. These consderatons motvate the hypothess: H3 (The Party Manfestos Hypothess): Ctzens ssue prortes and ther party support reflect partes ssue emphases, as artculated n ther electon manfestos. Below we report emprcal analyses of our three hypotheses n the context of Germany, an deal test case gven that the German panel survey data we analyze ncludes questons pertanng to many dmensons of ctzens ssue concerns. The multparty and proportonal character of the German party system allows us to further generalze ssue ownershp theory, whch has been tested almost exclusvely n two-party, majortaran systems such as the US and Brtan. However n supplementary analyses usng Brtsh panel data we report results that support substantve conclu-

12 8 sons that are dentcal to those we report for Germany. Ths suggests that our fndngs may apply generally across Western European electorates. The German Party System and Issue Emphases We evaluate our three hypotheses n the context of German poltcs, for both theoretcal and practcal reasons. Theoretcally, Germany s an approprate settng because t features a promnent green party whch allows us to evaluate our green party exceptonalsm hypothess, and moreover the multparty and proportonal character of German poltcs allows us to evaluate mportant aspects of ssue ownershp theory outsde the majortaran contexts of the US and Brtan, where t has prevously been tested. Practcally, Germany and Brtan are the only western European poltes for whch long-term panel survey data s avalable that ncludes detaled questons about respondents ssue prortes and ther party support, whch we requre n order to test our hypotheses. As we noted n the ntroducton, we have replcated our analyses on the Brtsh data, and these analyses support the same substantve conclusons as our German analyses. Snce the establshment of the West German democratc state n 1949, the German system has featured four major partes. The Chrstan Democrats (CDU/CSU; hereafter the CDU) 2 are a large, moderate, manstream party that supports busness-frendly, free-market economc polces, prortzes economc growth over envronmental protecton, emphaszes law and order ssues, and presents conservatve postons on socal ssues along wth a skeptcal atttude towards mmgraton and multculturalsm (see, e.g. Pardos-Pardo et al. 2013). The Free Democrats (FDP) are a smaller market-lberal party whch, lke the CDU, advocates pro-busness polces and whch served as a junor partner n coalton government wth the CDU between , , 2 The Chrstan Democratc Unon (CDU) and the Chrstan Socal Unon (CSU) can be consdered one party, also formng one facton n parlament. The latter competes n the federal state of Bavara. Hereafter we only refer to the CDU, whch ncludes CSU partsans.

13 , and The major dfferences between the FDP and the CDU are that the FDP s even more strongly pro-busness than the CDU, whle de-emphaszng law and order, multculturalsm, and socal ssues. The two major leftst West German partes over the past thrty years are the Socal Democratc Party (the SPD) and the Greens, who formed an alternatve leftst Coalton that was n government between 1998 and The SPD s a large, moderate, center-left party that typcally supports expandng socal welfare programs, that takes a mxed poston on the trade-off between prortzng the economy versus the envronment (Benot and Laver 2006), and that typcally de-emphaszes law and order ssues compared to ts rght-wng compettors. Fnally, the Greens strongly emphasze envronmental protecton, de-emphasze economc growth and law and order ssues, and also stake out more postve stances on multculturalsm than do the manstream partes. < Fgure 1 about here > Fgure 1 dsplays data collected by the Comparatve Manfesto Project (CMP; Budge et al. 2001) 3, whch plots the proportons of quas-sentences n each party s electon manfesto devoted to economc ssues (Fgure 1A), the envronment (Fgure 1B), law and order (Fgure 1C), and the sum of the party s negatve references to multculturalsm and ther postve references to the na- 3 The updated data avalable at was used for these analyses. To measure economc salence, we add the CMP tems , whch measure dverse aspects of the natonal economy such as free market economcs, economy plannng, and economc growth. Envronmental salency s measured usng CMP tem 501, whch ncludes postve mentons of envronmental protecton. CMP tem 605 s used to measure postve mentons of law and order. The sum of CMP tems 601 (Natonal Way of Lfe: Postve) and 608 (Multculturalsm: Negatve) s our measure of the salency of ant-mmgrant sentments. We follow Pardos-Prado et al. (2013) n conceptualzng mmgraton as an ssue salence rather than a postonal ssue.

14 10 tonal way of lfe, whch appear relevant to mmgraton ssues 4 (Fgure 1D) over each electon held between 1983 and 2009, the tme perod covered n the ndvdual-level analyses presented below. Fgure 1A documents that the manstream partes n partcular the CDU and the FDP strongly emphaszed economc ssues, compared to the Greens. As expected, the patterns n Fgure 1B demonstrate that the Greens dsproportonately emphaszed envronmental ssues, followed by the SPD. Fgures 1C and 1D document that the CDU dsproportonately emphaszed law and order ssues whle talkng more negatvely about multculturalsm (and postvely about the natonal way of lfe). These patterns across the dfferent ssue domans comport well wth experts understandng of German partes long-term ssue emphases. 5 Emprcal Analyss We evaluate our hypotheses va analyses of data from a unque 26-wave German panel study, the German Soco-Economc Panel (SOEP), whch tracks ctzens party support and ssue salence between 1984 and 2009 through annual face-to-face ntervews. 6 The study contans varous samples, such as separate Eastern German and refreshment samples; however we lmt our analyss to West German ctzens. We exclude East Germans and mmgrants as the nature of partsanshp and poltcal atttudes dffers for these groups due to dfferent socalzaton experences (Neundorf 2009; Kroh 2014). We analyze 19,777 respondents wth at least three observatons on 4 We note that the Comparatve Manfesto Project codngs do not nclude codngs of mmgraton. whch s why we analyze codngs for multculturalsm and natonal way of lfe. 5 These codngs are consstent wth surveys conducted by Benot and Laver (2006), where poltcal experts were asked to evaluate the relatve emphases that partes placed on envronmental protecton versus economc growth. 6 For more nformaton on the SOEP contents and structure see Hasken-DeNew and Frck (2005) and Wagner et al. (2007).

15 11 the party support and ssue prorty varables, the mnmum number requred to estmate our models. 7 We conducted supplementary analyses usng hgher cut-off ponts whch supported substantve conclusons dentcal to those we report below. The key varables n our analyses pertan to respondents partsanshp and ssue prortes. The partsanshp queston reads: Many people n Germany are nclned to a certan poltcal party, although from tme to tme they vote for another poltcal party. What about you: Are you nclned generally speakng to a partcular party? Those who responded, Yes, were then asked, Whch one? and handed a card that lsted all the partes. Those gvng no answer or don t know were set to mssng. The dependent varable was measured by dstngushng the supporters of the four major partes the SPD, CDU, FDP, and the Greens from ndependents and partsans of smaller partes. Issue salence was measured by the degree of concern respondents expressed wth respect to a seres of ssues. The queston wordng was: What about the followng areas: Are you concerned about them? Very concerned; 2. Somewhat concerned; 3. Not concerned at all. We beleve that the statement that somebody s very concerned denotes that the respondent prortzes the ssue, and we dchotomze the ssue concern varable accordngly. The word concern was already used n Petrock s semnal artcle (1996: 826) to capture ndvdual-level ssue salency (see also Pardos-Prado et al. 2013). We analyzed the ssues that are most appealng for evaluatng our hypotheses, namely respondents concerns over envronmental protecton, general economc development, crme, and mmgraton, whch we see as closely related to concerns over 7 We restrct the estmaton to respondents wth at least three vald responses, as ths provdes at least two changes n atttudes and/or partsanshp per person. Ths s needed to dentfy the effect correctly, as otherwse the estmaton s based on one change only, whch could have been randomly postve, negatve or non-sgnfcant. Only wth at least two of those changes s t possble to dentfy the drecton of the effect (e.g. ssue cung). For more nformaton, see also Neundorf et al. (2011).

16 12 multculturalsm and the natonal way of lfe (the relevant ssue domans that are ncluded n the CMP codngs). 8 The ssue concern questons pertanng to the economy and the envronment were asked across all 26 waves of the SOEP survey, and those pertanng to crme and mmgraton were asked between 1999 and Explorng partsans ssue salence Table 1 reports the proportons of respondents who stated that they were concerned wth the four polcy ssues, stratfed by party support. We see that partsans ssue concerns reflected ther partes manfesto-based ssue emphases, n that, for example, the proporton of Green supporters who expressed envronmental concerns (62.2%) sharply exceeded the correspondng proportons for manstream partes supporters and for ndependents (none of these groups exceeded 43%), whle Green partsans expressed far less concern over the economy, crme, and mmgraton than dd other respondents. Ths supports the vew of the Greens as a sngle-ssue nche party that owns the envronmental ssue, a pattern consstent wth the Green Party exceptonalsm hypothess. In addton, consstent wth the CDU s manfesto-based emphases, CDU supporters expressed the most concern wth crme and mmgraton. < Table 1 about here > Whle the fgures n Table 1 are suggestve, they do not allow us test the causal order of ssue salence and partsanshp. Do ctzens take ssue prorty cues from ther preferred party, or are ctzens party evaluatons drven by ther pre-exstng ssue prortes? Moreover, how do 8 The ssues are moderately correlated at 0.14 (envronment and mmgraton) to 0.30 (economy and crme). However, because mmgraton and crme are correlated at 0.50 we decded to estmate separate models for each ssue to avod mult-collnearty ssues. However we note that we also estmated models ncludng all ssues smultaneously, and these estmates supported substantve conclusons that were dentcal to those we report below.

17 13 these lnks connect to the ssues each party emphaszes n ts polcy manfestos? Below we present longtudnal analyses that address these questons. Statstcal specfcatons usng cross-lagged Markov Chan models To evaluate our hypotheses we model the dynamcs of SOEP respondents party support and ssue salence usng cross-lagged Markov Chan modelng, whch allows the consderaton of autocorrelaton n repeated observatons as well as the ncluson of lagged tme-varyng effects of ssue salence on partsanshp, and vce versa. Markov models employ a frst-order markovan structure allowng sequences of ndvdual observatons to be correlated, and recent studes by Clarke and McCutcheon (2009) and Neundorf et al. (2011) demonstrate that Markov models correctly specfy the dynamcs of ndvdual-level partsanshp. Specfyng the mpact of respondents lagged ssue concerns on partsanshp (partsan updatng effects). We model party support va a seres of multnomal logt equatons. Specfcally, the probablty that a respondent states that she s a partsan of party k at tme t, relatve to the probablty that s classfed as an ndependent, s estmated as a functon of overall ntercepts, s reported partsanshp at the prevous panel wave at tme t 1 (the effect of whch s allowed to vary across tme 9 ), and s expressed concerns wth the economy, the envronment, crme, or mmgraton at tme (t 1). We estmated four dfferent models to nclude one ssue at a tme. In the case of West Germany, whch features four major partes and the example of envronmental concerns, ths model s specfed as follows: 10 9 We allow these effects to vary across tme because party support s nfluenced by tme-specfc events such as poltcal scandals and crses that enhance (or depress) partes popular appeal. For nstance n 1999 the German meda exposed the llegal campagn donatons that the CDU had prevously accepted under the leadershp of Chancellor Helmut Kohl, a story lne that badly damaged the CDU s mage (Papp, Shkano, and Bytzek 2004). 10 The Baum-Welch algorthm mplemented n the Syntax verson of LatentGOLD (Vermunt and

18 14 P( Greens ( ) 1) t log 0G 1 GtGreens ( t 1) 2Gt SPD ( t 1) 3 P( Nonpartsa n ( t) 1) FDP ( t 1) envronmen t ( t 1) 4Gt 5G Gt CDU ( t 1), (1) P( SPD ( ) 1) t log 0S 1StGreens ( t 1) 2St SPD ( t 1) 3StCDU P( Nonpartsa n ( t) 1) FDP ( t 1) envronmen t ( t 1) 4St 5S ( t 1), (2) P( CDU ( ) 1) t log 0C 1 CtGreens ( t 1) 2Ct SPD ( t 1) 3 P( Nonpartsa n ( t) 1) FDP ( t 1) envronment ( t 1) 4Ct 5C Ct CDU ( t 1), (3) P( FDP ( ) 1) t log 0F 1FtGreens ( t 1) 2Ft SPD ( t 1) 3 P( Nonpartsa n ( t) 1) FDP ( t 1) envronmen t ( t 1) 4Ft 5F Ft CDU ( t 1). (4) In equaton 1,, denote stablty coeffcents whch denote how s 1 Gt 2Gt 3Gt, 4Gt lagged partsanshp affects her current lkelhood of supportng the Greens (relatve to her lkelhood of beng ndependent), where lagged partsanshp s specfed by the dummy varables Greens ( t 1), SPD ( t 1), CDU( t 1), FDP ( t 1) whch equal one f supported the focal party at tme (t 1) and zero otherwse. Of course we expect that respondents who supported the Greens at tme (t 1) are lkely to support the Greens at tme t,.e., we expect a postve coeffcent estmate 1Gt. Note that we also estmate effects on Green Party support of respondents lagged support for the SPD, CDU, and FDP to evaluate whether dfferent partes eltes provde dfferng cues wth respect to the Greens. For nstance we mght expect SPD eltes to cue ther supporters to postvely evaluate the Greens, gven these partes hstory of collaboraton n natonal government. Magdson 2008) was used to handle the large number of cases n our panel study. 25 start sets per model were estmated. The fnal set of parameters were estmated after 1000 EM teratons usng the Newton s methods.

19 15 The coeffcent 5G n equaton 1 denotes the mpact of s lagged envronmental concerns represented by the dummy varable envronment(t 1) on s partsanshp at tme t. A postve (negatve) cross-lagged coeffcent estmate on 5G denotes that s lagged envronmental concerns enhance (depress) her lkelhood of currently supportng the Greens, whch would be evdence of a partsan updatng effect wth respect to the Greens. Specfyng the mpact of respondents lagged partsanshp on ther ssue concerns (ssue cueng effects). We specfy the probablty that a respondent s concerned wth a focal ssue at tme t as a functon of overall ntercepts, her ssue concerns at the prevous tme (t 1) as tme-varyng perod effects, and her lagged partsanshp. Below we present the specfcaton for ctzens envronmental concerns; the specfcatons for the remanng ssues (the economy, crme, and mmgraton) dsplay the same functonal form: P( envronment ( ) 1) t log 0 1 tenvronment ( t 1) 2Greens ( t 1) P( envronment ( t) 0) SPD ( t 1) CDU ( t 1) FDP ( t 1) (5) In equaton 5 the coeffcent 1t denotes a stablty coeffcent that nfluence respondents envronmental concerns 11 whle the cross-lagged coeffcents 2, 3, 4, 5 capture the partsan updatng effects of lagged partsanshp. Thus a postve estmate on 2, the coeffcent on the Greens (t 1) varable, wll denote that respondents who supported the Greens at tme (t 1) were more lkely to express envronmental concerns at tme t, when controllng for lagged envronmental concerns an estmate that would mply that the Greens cue ther supporters to pror- 11 Such tme-specfc effects nclude events such as envronmental dsasters (such as the Chernobyl nuclear accdent), whch depress or enhance respondents envronmental concerns ndependently of ther partsanshp.

20 16 tze the envronment. The coeffcents 3, 4, 5 represent parallel estmates of envronmental cues assocated wth lagged support for the SPD, CDU, and FDP. Control varables on ntal partsanshp and ssue concerns. Fnally, our specfcatons ncluded ndvdual-level covarates to account for factors that affected respondents partsanshp and ssue prortes when they frst entered the panel (see Neundorf et al. 2011). We expect that educaton, occupaton, age, gender, church attendance and poltcal nterest affect respondents ntal partsanshp and ssue prortes. For example, the results show that those who are poltcally nterested are more lkely to be partsans and to be concerned about poltcal ssues. In supplementary materals we report the coeffcent estmates on these varables. Results The recprocal mpact of ssue salence and partsanshp. The upper panels (grey bars) of Fgures 2A-2D dsplay the estmated logt coeffcents for equatons 1-5 above, along wth the 95% confdence ntervals on these estmates. 12 The dark-grey bars represent partsan updatng effects of respondents lagged ssue salence on ther current party support for the four ssue areas we analyze, whle the lght-grey bars dsplay the coeffcents of lagged partsanshp on current ssue prortes,.e., ssue cueng effects. < Fgure 2 about here > The results dsplayed n the upper panels of Fgures 2A-2D support the recprocal effects hypothess, that ctzens ssue salence both nfluence and are nfluenced by ther party support. The estmates n Fgure 2B denote that, holdng lagged partsanshp constant, respondents wth lagged envronmental concerns were more lkely to support the Greens and the SPD and less 12 For the nterested reader, we also report the table ncludng the numerc expresson of these coeffcents n the Appendx Table A1.

21 17 lkely to support the CDU and the FDP at the current panel,.e., the coeffcent estmates on lagged envronmental concerns are postve wth respect to the Greens and the SPD and negatve for the CDU and the FDP (p <.01 n all cases). Recpro-cally, we estmate that lagged support for the Greens and the SPD cued respondents to pro-rtze the envronment at the current panel, whle lagged CDU and FDP support cued resp-ondents to de-emphasze ths ssue (p <.01). Wth respect to the economy (Fgure 2A) we estmate that, ceterus parbus, lagged support for the CDU and the SPD cued respondents to prortze ths ssue at the current panel, whle lagged Green support cued respondents to de-emphasze the economy (all coeffcents sgnfcant at p <.01). Fnally, Fgures 2C-2D dsplay estmates that lagged crme and mmgraton concerns prompted respondents to wth-draw support from the FDP, the SPD, and the Greens at the current panel (p <.01), and, rec-procally, that respondents who reported lagged support for these partes deemphaszed these ssues at the current panel (p <.01). These estmates support the recprocal effects hypothess. Fgures 2A-2D allow us to drectly compare the sze of the recprocal effects and to assess whch effect s stronger. Despte sgnfcant cross-lagged effects n both drectons whch supports the recprocal effects hypothess the ssue cung effects are generally larger than the partsan updatng effects. We see that especally for the Green Party, Green partsanshp strongly cues voters to be concerned about the envronment whle depressng ther concerns about the economy, crme, and mmgraton. Ths supports the Greens profle as a sngle-ssue nche party. We also estmate stronger ssue cung effects for SPD supporters for the ssues of the economy and the envronment, compared to the recprocal partsan updatng estmates In supplementary materals we further report the model ft, calculated as the dfference n AIC between the model excludng a cross-lagged effect of ssue salency (at t-1) on partsanshp and vce versa, and the model ncludng these cross-lagged coeffcents. The models ncludng partsanshp when predctng ssue salency clearly outperform the mprovement of the model compared to the predcton of the partsan updatng effect.

22 18 The queston arses: Can we reasonably nfer causal relatonshps from the statstcal assocatons we estmate,.e., that ctzens party support and ther ssue concerns recprocally nfluence each other? Whle we cannot defntvely resolve ths ssue, we see strong reasons to nfer causal relatonshps. Wth respect to ssue cueng effects, for nstance, we uncover strong assocatons between survey respondents lagged party support and ther current ssue concerns, even when controllng for respondents lagged party support. Gven the strong theoretcal reasons we have outlned to expect ctzens to take ssue-based cues from poltcal partes, and gven that we also control for respondent characterstcs ncludng educaton, occupaton, age, gender, church attendance, and poltcal nterest thereby controllng for (at least some of) the factors that mght jontly nfluence ctzens ssue prortes and ther party support we nfer that ctzens party support ndeed exerts a causal nfluence on ther ssue concerns. Ths nference s strengthened by the emprcal analyses we report below, whch drectly lnk the ndvdual-level ssue cueng processes we estmate from the panel data to the ssues the partes emphasze n ther electon manfestos. Thus, whle our analyses cannot defntvely prove a causal relatonshp a caveat that apples to most cross-lagged analyses of panel survey data we nfer that ctzens ssue concerns and ther party support do ndeed recprocally nfluence each other, and we employ causal language n the remander of ths paper. In ths sprt we proceed. Issue salence, partsanshp and Green Party exceptonalsm The estmates dsplayed n Fgure 2 support the Green Party exceptonalsm hypothess, that ssue-based effects are far stronger wth respect to the Greens than for manstream partes. Specfcally, for all four ssue areas the coeffcent estmates on the Greens wth respect to both partsan updatng and ssue cueng are over three tmes the magntudes of the estmates on any manstream party. (In all cases the dfferences between the estmates on the Greens versus manstream partes are statstcally sgnfcant, p <.01.) Ths strkng dfference suggests that mass-elte lnkages nvolvng the Greens dffer fundamentally from those nvolvng the manstream partes. Smply put, German ctzens ssue prortes strongly nfluence and are

23 19 nfluenced by ther support for the Greens, whle the parallel effects wth respect to manstream partes are modest. And, we emphasze that ths pattern extends to every ssue we examne, not only the envronment where we fnd as expected that envronmental concerns push ctzens towards the Greens (and vce versa): we also estmate that lagged concerns over the economy, crms and the envronment drve ctzens sharply away from the Greens to a much greater extant than such concerns push ctzens towards (or away from) any manstream party and that lagged Green Party support sharply depresses respondents lkelhoods of prortzng these ssues. Lnkng ndvdual-level ssue effects to party manfestos Next, we evaluate whether the ssue-based effects we estmate correspond wth the ssue emphases the partes presented n ther electon manfestos. The lower panels of Fgures 2A-2D dsplay the partes long-term ssue emphases (averaged over the perod ) based on the CMP manfesto codngs, presented earler n Fgure 1. These party-level ssue emphases strongly correlate wth the ndvdual-level partsan updatng and ssue cueng effects we estmate from the German panel data. For example the Greens, followed by the SPD, most strongly emphasze envronmental ssues n ther electon manfestos (see the lower panel of Fgure 2B), and we estmate strongly postve ndvdual-level partsan updatng and ssue cueng effects on ths ssue wth respect to these two partes,.e., that lagged envronmental concerns enhance ctzens support for the Greens and the FDP, and that lagged support for these partes prompts ctzens to prortze the envronment (ceterus parbus). Meanwhle we estmate negatve ndvdual-level partsan updatng and ssue cueng effects on the envronment wth respect to the CDU and the FDP, the two partes that de-emphasze ths ssue n ther manfestos (see the lower panel of Fgure 2B). Overall, the correlaton between the partes manfesto-based envronmental emphases and our estmates of ndvdual-level partsan updatng effects for each party as well as our ssue cueng estmates s 0 98 (p =.001). These strong assocatons extend to the remanng ssues: The CDU places the strongest manfesto-based emphass on crme and mmgraton (see the bottom panels

24 20 of Fgures 2C-2D), and t s the only party for whch we estmate postve ndvdual-level partsan updatng and ssue cueng effects on these ssues, whle we estmate strongly negatve ndvdual-level effects wth respect to the Greens, the party whch devotes the least attenton to these ssues. The correlaton between the partes manfesto-based crme emphases and our estmates of ndvdual-level partsan updatng effects s 0.90 (p =.001), whle the correlaton between the partes crme emphases and our estmates of ndvdual-level ssue cueng effects s 0.94 (p =.001); and, the correlatons on the mmgraton ssue are 0.74 (p =.001) for ndvdual-level partsan updatng effects, and 0.75 (p =.001) for ssue cueng effects. Fnally, wth respect to economc ssues, the correlaton between the partes manfesto-based economc emphases are our estmates of ndvdual-level partsan updatng effects s 0.71 (p =.01) and the correlaton between the partes manfesto-based economc emphases and our estmated ssue cueng effects s 0.63 (p =.01). These strong assocatons support the party manfesto hypothess that partes ssue emphases, as artculated n ther electon manfestos, strongly nfluence ctzens ssue prortes and ther partsanshp. Illustratng the recprocal effect of partsanshp and ssue salence Fgure 3 dsplays the substantve mpact of our estmated ssue cueng effects. 14 The fgures plot the probabltes that a respondent who dd not prortze the focal ssue (the economy, the envronment, crme, or mmgraton) at the prevous panel wave would prortze ths ssue at the cur- 14 For comparson purposes, n supplementary materals we report analyses pertanng to how lagged ssue concerns (the economy, the envronment, crme, and mmgraton) nfluence current party support. These analyses support the same substantve conclusons as the analyses we present here, on how lagged partsanshp nfluence the ssue prortes of respondents who dd not express these ssue concerns n the prevous panel wave. We have further re-produced the graphs presented n Fgure 3 for the stablty of partsanshp and ssue salence, whch are avalable n supplementary materals.

25 21 rent panel wave, as a functon of her lagged partsanshp. To better dstngush the strength of these effects, we further nclude the mean ssue moblzaton among ndependents as a baselne. < Fgure 3 about here > Consstent wth the Green Party exceptonalsm hypothess, we see that lagged Green Party support strongly cued respondents current ssue prortes. Wth respect to envronmental ssues, Fgure 3B dsplays results that among respondents who dd not report lagged envronmental concerns, those who were poltcal ndependents at the prevous panel had a computed 22.2% probablty of prortzng the envronment at the current panel wave, whle for lagged Green Party supporters ths computed probablty jumped to 41.3%, nearly double that for ndependents. The Fgure also dsplays how strongly the Greens cued ther supporters to de-emphasze other ssue areas: Fgures 3A, 3C, and 3D dsplay computatons that among respondents who dd not prortze the economy, crme, and mmgraton at the prevous panel wave, lagged ndependents had computed probabltes of 21.1%, 23.8%, and 16.4%, respectvely, of prortzng these ssues at the current panel, whle lagged Green Party supporters computed probabltes of prortzng these ssues were only 16.4%, 11.3%, and 4.3%. These computatons mply that the Green Party strongly cued ts supporters attenton towards the envronment, and away from all other ssues. Ths suggests that the dramatc dfferences n Green Party supporters ssue prortes vs-à-vs manstream partsans prortes, presented earler n Table 1, reflect n part the strong ssue prorty cues that Green Party supporters take from ths party s eltes. And ths n turn mples that the close match between the Green Party s ssue prortes, as reflected n ts electon manfestos, and ts supporters ssue prortes, reflects n part he party s ablty to shape ts supporters prortes. Fgures 3A-3D also dsplay computatons that reflect the manstream partes (more modest) abltes to shape ther supporters ssue concerns. On crme and mmgraton, for nstance, Fgures 3C-3D llustrate that lagged support for the CDU the party that most strongly hghlghts these ssue n ts manfestos ncreases respondents lkelhoods of prortzng crme and mm-

26 22 graton at the current panel wave by two to four percentage ponts (compared to lagged ndependence), whle lagged support for the SPD and the FDP decreases respondents lkelhoods of prortzng these ssues by two to sx percentage ponts. Wth respect to the envronment, Fgure 3B llustrates that lagged support for the SPD whch emphaszed envronmental ssues more strongly than the CDU and FDP (see the bottom panel of Fgure 2D) ncreased respondents lkelhood of prortzng envronmental ssues by about three percentage ponts (agan compared to the baselne of lagged ndependence), whle lagged support for the CDU and the FDP depressed the lkelhood of subsequent envronmental concerns by two to three percentage ponts. 15 Robustness checks We conducted addtonal analyses on the German data along wth extensons to Brtsh panel data to assess the robustness of our conclusons. Wth respect to Germany, we analyzed whether our conclusons vared dependng on whch partes were currently n the natonal governng coalton, and we also estmated the parameters of models that specfed longer tme lags for the recprocal relatonshps between respondents ssue prortes and partsanshp, compared to the one-year lagged specfcatons reported above. In addton, to nvestgate the effects of possble measurement error we replcated our models whle specfyng partsanshp as a latent varable. Lastly, we condtoned the cross-lagged effects on the age of the respondents to test the possblty that as ctzens age these cross-lagged effects dmnsh,.e., that older respondents ncreasngly resst updatng ther partsanshp and ther ssue prortes. These analyses contnue to support our substantve conclusons: we fnd that mass-elte ssue lnkages vared only modest- 15 For comparson purposes, n supplementary materals we report analyses pertanng to how lagged concerns about an ssue (the economy, the envronment, crme, and mmgraton) nfluence current partsanshp. These analyses support the same substantve conclusons as the analyses we present here, on how lagged partsanshp nfluence the ssue prortes of respondents who dd not express these ssue concerns n the prevous panel wave.